MixMC: A Multivariate Statistical Framework to Gain Insight into Microbial Communities

Abstract

Culture independent techniques, such as shotgun metagenomics and 16S rRNA amplicon sequencing have dramatically changed the way we can examine microbial communities. Recently, changes in microbial community structure and dynamics have been associated with a growing list of human diseases. The identification and comparison of bacteria driving those changes requires the development of sound statistical tools, especially if microbial biomarkers are to be used in a clinical setting. We present , a novel multivariate data analysis framework for metagenomic biomarker discovery. accounts for the compositional nature of 16S data and enables detection of subtle differences when high inter-subject variability is present due to microbial sampling performed repeatedly on the same subjects, but in multiple habitats. Through data dimension reduction the multivariate methods provide insightful graphical visualisations to characterise each type of environment in a detailed manner. We applied to 16S microbiome studies focusing on multiple body sites in healthy individuals, compared our results with existing statistical tools and illustrated added value of using multivariate methodologies to fully characterise and compare microbial communities.

mixOmics: an R package for ‘omics feature selection and multiple data integration

“…Such an analysis suggests novel biological hypotheses to be further validated in the laboratory, when one is seeking for a signature of a subset of features to explain, discriminate or predict a categorical outcome. The methods has been applied and validated in several biological and biomedical studies, including ours in proteomics and microbiome Shah et al (2016);Lê Cao et al (2016).…”

“…While mixOmics methods can handle large data sets (several tens of thousands of predictors), we recommend pre-filtering the data to less than 10K predictors per data set, for example by using Median Absolute Deviation Teng et al (2016) for RNA-seq data, by removing consistently low counts in microbiome data sets Arumugam et al (2011);Lê Cao et al (2016) or by removing near zero variance predictors. Such step aims to lessen the computational time during the parameter tuning process.…”

HLA Alleles Associated With Risk of Ankylosing Spondylitis and Rheumatoid Arthritis Influence the Gut Microbiome

“…Multidimensional data visualization was conducted using sparse partial least squares–discriminant analysis (sPLSDA) on centered log‐ratio–transformed data, as implemented in R as part of the MixOmics package, version 6.3.1 . Association of the microbial composition with metadata of interest was conducted using a permutational multivariate analysis of variance (PERMANOVA) test as part of R‐vegan, version 2.4‐5 , on arcsine square root–transformed data at species level.…”

Shotgun metagenomics reveals an enrichment of potentially cross-reactive bacterial epitopes in ankylosing spondylitis patients, as well as the effects of TNFi therapy and the host’s genotype upon microbiome composition

“…Abundance tables were arcsine square root transformed prior to analysis. Multidimensional data visualisation was conducted using a sparse partial least squares discriminant analysis (sPLSDA) as implemented in R as part of the MixOmics package v6.3.1 38 , at the species level using Bray-Curtis distance matrices. Receiver operating characteristic curve was calculated from sPLSDA using the MixOmics package v6.3.1.…”